Abe Hou

PhD student at @stanfordnlp @stanford. Undergrad at @jhuclsp.

[John Yang @jyangballin]

Had the best time chatting w/ @vincentsunnchen! Shared some thoughts around how far we've come and how far we might go in AI coding land. Hope you enjoy!

[Vishakh Padmakumar @vishakh_pk]

People are increasingly worried that AI tools make us overreliant.
But how do we actually measure this? We introduce Offloading Score, a measure of reliance based on the fraction of cognitive effort offloaded to AI while completing a task.
In a controlled user study, Offloading Score detects increased reliance under time pressure, while several common alternatives do not.
(1/9)

[Abe Hou @abe_hou]

Exciting workshop!

[Hao Zhu @_Hao_Zhu]

We are hosting a new COLM workshop on Agent Behavior on 10/9 in SF

We invite two kinds of submissions at this stage: papers and benchmark proposals (super lightweight, we will provide credits for running your benchmark on frontier models).

Check out Manuel’s post and our website!

[Hao Zhu @_Hao_Zhu]

> If we are to realize the promised future where AI agents work together – and with human collaborators – AI will have to be good collaborators, but so far they appear to be lone wolves

Our research on the curse of coordination in agent teams featured in @StanfordHAI blog
https://t.co/fsbZq1QWMc

[Augmented Mind Podcast @augmind_fm]

The AM Podcast EP5 is around the corner. 👀🧑‍🍳

You may have seen the viral post about OpenAI math breakthrough. For EP5 we sat down with Jeremy Avigad, co-creator of @leanprover before that and the conversation is incredible. Jeremy shares not just the Lean project, but also
- The verification gap in human-AI collaboration
- How AI is changing mathematics
- The future of math education
- Capital, startups, and the mathematician's ecosystem
- ...

Stay tuned and subscribe our YouTube channel for latest updates!

[Augmented Mind Podcast @augmind_fm]

Join us this Thursday May 28th 6-7pm PST for our first ever AM Podcast Live Stream! 🎉

We are hosting @cjziems, @dorazhao9, and @Diyi_Yang for a discussion on their new paper "Reflections and New Directions for Human-Centered Large Language Models"!

RSVP 🔗⬇️

[Dora Zhao @dorazhao9]

Raise your hand if you have heard about "human-centered" AI or LLMs and not been sure what that means? 🙋‍♀️🙋‍♀️🙋‍♀️

Check out our paper for answers :)

[Diyi Yang @Diyi_Yang]

The next frontier of AI is not only more capable model; it is an AI that *humans* can meaningfully live and work with :)

With all students in my cs329x Human-Centered LLM class, we present 60+ pages of insights for developing Human-Centered LLMs (HCLLMs), from design & data sourcing to training, eval & deployment 🧵

[Yijia Shao @EchoShao8899]

Tomorrow (5/19), 6–7pm PT: We're going live for the first time to share how to collaborate with AI agents more effectively 🤖
Thrilled (and a little nervous!) to see so many RSVPs. As Stanford Qualtrics console is down😅, we can't reach everyone directly — here's your calendar link: https://t.co/bf4pbb9FsO

[Lujain Ibrahim @lujainmibrahim]

New preprint!

In 5 studies (3k+ users / 12k+ convs, with a 3-wk longitudinal study), we find that sycophantic AI influences how people view those closest to them.

It affects how effortful human interaction seems, how satisfying it is, & who people want to turn to for advice 🧵

[Changyu Chen @Cameron_Chann]

Life update: I'm super excited to join @Stanford as a postdoc working with @Diyi_Yang ! I’ll continue my research on RL, and recently I’ve become especially interested in how RL can contribute to human-AI collaboration and collaborative agents.

A new chapter begins, from the sunny island to the sunny state ☀️🏝️

[Houjun Liu @houjun_liu]

🚨 Your coding agent may be secretly sticking vulnerabilities into your code!! 🚨

Wouldn't you want to fix that? Hint: asking it to write secure code is not enough. (1/n)

[Abe Hou @abe_hou]

@michaelryan207 @KnightHennessy Huge congrats Michael! 🐐💕

[Michael Ryan @michaelryan207]

Super excited to join the 2026 cohort of @KnightHennessy scholars!

Has been incredible talking to the other scholars about the major problems they are tackling across healthcare, science, policy, etc. Excited to work towards Human-Centered Open-Source AI that can support people tackling the world's biggest challenges!

[Fang Wu @WUFang40615703]

Not all diffusion noise is equally useful for training!🤫

We introduce 🐯NoiseRater: a meta-learned framework that scores and selects informative noise instances during diffusion training.

Instead of treating Gaussian noise uniformly, we learn which noise samples actually improve downstream generalization.

Results:
• Better FID on FFHQ + ImageNet
• Improved training efficiency
• Transfers across DiT backbone sizes

A new axis for improving diffusion training: training-time noise valuation.
Paper: https://t.co/xbCJlTTEA4

Great thanks to collaborators @Hanqun_CAO @XiangruTang @hcwww_ @MoleiTaoMath @KKuanPang @erranlli and my advisors @YejinChoinka @jure

#AI #MachineLearning #DiffusionModels

[Michael Ryan @michaelryan207]

Just released EP4 of the AM Podcast (@augmind_fm)! 🎙

I had a ton of fun talking with Ken Liu @kenziyuliu, creator of the Open Anonymity Project!  Ken is an expert in both privacy technologies and AI, and there isn’t anyone I would trust more to keep my AI conversations private and secure.🔒

Listen for a deep dive into the algorithms that power Open Anonymity’s privacy guarantees and for why Ken believes LLMs create a uniquely concerning privacy threat.

[Omri Uzan @omri_uzan]

Can we improve a black-box retriever’s performance without additional query-time latency?

We introduce Document Optimization: documents are rewritten offline into retrieval-optimized surrogates by a language model. The LM/VLM is trained with RL, with rewards derived from the retriever's rankings.

Work with Ron Polonsky, @douwekiela and @ChrisGPotts

👇🧵

[Fang Wu @WUFang40615703]

🚨 4 papers accepted to ICML 2026 🚨

Here’s a quick thread 🧵👇

🧬 1. Proteo-R1 — Thinking before generating molecules

Most generative models in biology are powerful but largely intuitive (pattern matching).

We ask: can they reason like scientists?

→ Introduce reasoning-guided diffusion
→ Embed step-by-step “thinking” directly into the generative trajectory
→ Turn molecule generation into an interpretable reasoning process

This bridges symbolic reasoning + physical generation and leads to strong gains in protein binder design

⚡ 2. BroRL — A different scaling law for RL

RLVR has been scaling via more training steps… but hits a plateau

We explore a different axis:
👉 scale the number of rollouts per example

From a probability mass perspective:
more exploration → more mass shifts toward correct solutions

Result:
📈 sustained gains beyond prior saturation points
Simple idea — surprisingly underexplored

🧩 3. RiboSphere — Tokenizing RNA structure

RNA is not text — it’s geometry

We combine:
→ vector quantization
→ flow matching
to learn discrete, SE(3)-invariant representations of RNA 3D structures

The model captures motif-level structure
and transfers well to inverse folding + RNA–ligand binding. Especially strong in low-data regimes

🧪 4. LatentChem — Do models need language to think?

Chain-of-thought forces reasoning into text but chemical reasoning isn’t naturally linguistic

We introduce:
👉 latent-space reasoning

→ decouple reasoning from language
→ let models think in continuous space

We observe emergent internalization +
⚡ ~10× speedup with improved accuracy vs CoT

More details soon — would love to hear thoughts/feedback

#ICML2026 #AI #MachineLearning #AIforScience #BioAI #Chemistry #R

[Abe Hou @abe_hou]

Can't wait to find out

[Augmented Mind Podcast @augmind_fm]

New episode of the AM podcast dropping soon!
In EP4, we sat down with @kenziyuliu, CS PhD student at @StanfordAILab and creator of The Open Anonymity Project, to talk about the privacy layer of personal intelligence.

Here's a preview 😃

[Weiyan Shi @shi_weiyan]

Verbalized Sampling is accepted at ICML! 🥳

It's wild to see research turn into so much impact: two books + a startup already 😂 The most magical part is a high schooler told me they were skeptical of VS at first because it's such a simple change; but once they tried it, the AI suddenly started to explain concepts in more diverse ways to help them learn better♥️

glad VS is making the world a bit more diverse 🧡💚💙